Methods

Study Area

This study uses a simulated landscape drawn from data approximating the communities and conditions within Great Smoky Mountains National Park. Great Smoky Mountains National Park is a 2,110 km² World Heritage Site and International Biosphere Reserve straddling the border between western North Carolina and eastern Tennessee. Great Smoky Mountains National Park serves as an ideal model for this study as most major ecosystems of the Southern Appalachians are represented, and the general topographic distribution of communities and tree species has previously been described (Whittaker 1956).

The Southern Appalachian Mountains, although not representative of all eastern forests, are unique because they represent one of the most biologically diverse regions of the world (SAMAB 1996). A complex system of physiography, environmental site conditions, adaptive life history characteristics, and disturbance history has created a distinctive vegetation structure (Elliott and others 1999). Due to this complexity, Southern Appalachian landscapes contain a variety of community types ranging from mesophytic hemlock-hardwood forests on moist valley floors to yellow pine woodlands on xeric ridges and from low-elevation temperate deciduous forests to high-elevation spruce-fir forests (Whittaker 1956, Stephenson and others 1993). Such high biodiversity areas have been thought by some to act as potential barriers to invasion because of increased competition and by others as at risk of invasion due to the higher potential for suitable habitat niches (Brown 2002, Brown and Peet 2003, Elton 1958, Kennedy and others 2002, Levine and D’Antonio 1999).

Model Description

LANDIS is a spatially explicit computer model designed to simulate forest succession and disturbance across broad spatial and temporal scales (He and Mladenoff 1999a, b; He and others 1996; He and others 1999a, b; Mladenoff and He 1999). Whereas LANDIS was originally developed to simulate disturbance and succession on glacial plains in the Upper Midwest (Mladenoff 2004), it has been successfully adapted for use in mountainous areas (He and others 2002a; Shifley and others 1998, 2000; Waldron and others, in press; Xu and others 2004).

LANDIS is raster-based, with tree species (max 30) simulated as the presence or absence of 10-year-age cohorts on each cell. At the site (cell) scale, LANDIS manages species life history data at 10-year time steps. Succession is individualistic and is based on dispersal, shade tolerance, and land type suitability. Disturbances that can be modeled include fire, wind, harvesting, and biological agents (insects, disease) (Sturtevant and others 2004a).

Fire in LANDIS is a hierarchical stochastic processes based on ignition, initiation, and spread (Yang and others 2004). Mortality from fire is a bottom-up process whereby low-intensity fires kill young, fire-intolerant species, whereas fires of higher intensity can kill larger trees and more fire-tolerant species (He and Mladenoff 1999a).

Biological disturbances in LANDIS 4.0 are modeled using the Biological Disturbance Agent (BDA) module. Biological disturbances are probabilistic at the site (cell) level. Each site is assigned a Site Vulnerability (SV) probability value that is checked against a uniform random number to determine if that site has been infected. Site vulnerability can be directly equated with the Site Resource Dominance (SRD) value that ranges from 0-1 and is based on species and species age. This value can also be modified by three variables to determine the impact on a given site—Modified Site Resource Dominance (SRDm), Neighborhood Resource Dominance (NRD), and the temporal scale of outbreaks. The functioning of these variables and of the BDA in general is described in detail in Sturtevant and others (2004b).

Simulation Methods

Density of Yellow Pine and Hemlock

We used LANDIS 4.0 to simulate forest dynamics on a 120-ha idealized landscape. The landscape was a 100 by 120 cell grid with a cell size of 10 m by 10 m, the smallest cell size recommended for use with LANDIS. Using this small cell size allowed us to operate at approximately the scale of the individual canopy tree, following the logic of gap models (cf. Botkin 1993). The landscape was divided into 18 individual land types arranged according to the mosaic chart used by Whittaker (1956) to depict the elevation and moisture gradients on the Great Smoky Mountains landscape. The land types are arranged in three rows and six columns. The rows represent (from bottom to top) low (400–915 m)-, middle (916–1370 m)-, and high (1371–2025 m)- elevation zones. The columns represent different topographic moisture classes. Moisture availability decreases from left to right, as follows: (1) coves and canyons; (2) flats, draws, and ravines; (3) sheltered slopes; (4) east to northwest facing slopes; (5) southeast to west facing slopes; and (6) ridges and peaks. Elevation also influences moisture availability. For example, a low-elevation ridgetop would have drier conditions than a mid-elevation ridgetop. Although the simulated landscape incorporates the full range of environments in the Great Smoky Mountains, our interest in this paper is only on the successional patterns for those land types under the greatest threat by SPB and HWA (Figure on the right). We present results for mid-elevation ridges and peaks (SPB) and mid-elevation flats, draws, and ravines (HWA) to illustrate the utility of the model in assessing insect threats.